Optimal periodic flexible policies for two-stage serial supply chains

In a two-stage serial supply chain, a periodic flexible policy (PF policy) allows the retailer to receive fixed orders that may depend on demand history in one period of the ordering cycle and order freely in other periods. Existing literature has shown that certain PF policies can significantly reduce the inefficiency in a decentralized supply chain. However, these works have mostly defined and implemented ad-hoc periodic flexible policies and have not attempted to identify the optimal periodic flexible policies. In this paper, we characterize the structure of the optimal PF policy using calculus of variations. In particular, we show that under the optimal PF policy, the retailer receives shipments either according to a state dependent capacitated policy or a state dependent order up to policy. Furthermore, we can approximate the retailer's optimal restricted ordering function by a piecewise linear function and show numerically that this approximation is near optimal.